EP1241662B1 - Reconnaissance de la parole avec compensation de distorsion convolutif et de bruit additif - Google Patents
Reconnaissance de la parole avec compensation de distorsion convolutif et de bruit additif Download PDFInfo
- Publication number
- EP1241662B1 EP1241662B1 EP02100251A EP02100251A EP1241662B1 EP 1241662 B1 EP1241662 B1 EP 1241662B1 EP 02100251 A EP02100251 A EP 02100251A EP 02100251 A EP02100251 A EP 02100251A EP 1241662 B1 EP1241662 B1 EP 1241662B1
- Authority
- EP
- European Patent Office
- Prior art keywords
- mean
- vector
- speech
- calculating
- noise
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 239000000654 additive Substances 0.000 title claims description 12
- 230000000996 additive effect Effects 0.000 title claims description 12
- 239000013598 vector Substances 0.000 claims description 49
- 238000000034 method Methods 0.000 claims description 23
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 description 9
- 230000006978 adaptation Effects 0.000 description 6
- 230000003595 spectral effect Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000007476 Maximum Likelihood Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 230000005534 acoustic noise Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/20—Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/14—Speech classification or search using statistical models, e.g. Hidden Markov Models [HMMs]
- G10L15/142—Hidden Markov Models [HMMs]
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
Definitions
- This invention relates to speech recognition and more particularly to compensation for both background noise and channel distortion.
- a speech recognizer trained with relatively a quiet office environment speech data and operating in a mobile environment may fail due to at least to the two distortion sources of back ground noise and microphone changes.
- the background noise may be from a computer fan, car engine, and/or road noise.
- the microphone changer may be due to the quality of the microphone, hand-held or hands-free and, a position to the mouth. In mobile application of speech recognition, both the microphone conditioner and background noise are subject to change.
- Cepstral Mean Normalization removes utterance mean and is a simple and effective way of dealing with convolutive distortion such as telephone channel distortion. See “Effectiveness of Linear Prediction Characteristics of the Speech Wave for Automatic Speaker Identification and Verification” of B. Atal in Journal of Acoustics Society of America, Vol. 55: 1304-1312, 1974.
- Spectral Subtraction reduces background noise in the feature space. See article “Suppression of Acoustic Noise in Speech Using Spectral Subtraction” of S.F. Boll in IEEE Transactions on Acoustics, Speech and Signal Processing, ASSP-27(2): 113-129, April 1979.
- Parallel Model Combination gives an approximation of speech models in noisy condition from noise-free speech models and noise estimates.
- Joint compensation of additional noise and convolutive noise can be achieved by the introduction of a channel model and a noise model.
- a spectral bias for additive noise and a cepstral bias for convolutive noise are introduced in an article by M. Afify, Y. Gong, and J. P. Haton. This article is entitled "A General Joint Additive and Convolutive Bias Compensation Approach Applied to noisy Lombard Speech Recognition" in IEEE Trans. on Speech and Audio Processing, 6(6): 524-538, November 1998.
- the five biases can be calculated by application of Expectation Maximization (EM) in both spectral and convolutive domains.
- EM Expectation Maximization
- Gauvain, et al is presented to calculate convolutive component, which requires rescanning of training data. See J.L. Gauvain, L. Lamel, M. Adda-Decker, and D. Matrouf entitled “Developments in Continuous Speech Dictation using the ARPA NAB News Task.” In Proc. of IEEE International Conference on Acoustics, Speech, and Signal Processing, pages 73-76, Detroit, 1996. Solution of the convolutive component by steepest descent method has also been reported. See Y. Minami and S. Furui entitled “A Maximum Likelihood Procedure for a Universal Adaptation Method Based on HMM Composition.” See Proc. of IEEE International Conference on Acoustics, Speech and Signal Processing, pages 129-132, Detroit, 1995. A method by Y.
- Minami and S. Furui needs additional universal speech models, and redestination of channel distortion with the universal models when channel changes. See Y. Minami and S. Furui entitled "Adaptation Method Based on HMM Composition and EM Algorithm" in Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing, pages 327-330, Atlanta 1996.
- the nonlinear changes of both type of distortions can be approximated by linear equations, assuming that the changes are small. Jacobian approach, which models speech model parameter changes as the product of a jacobian matrix and the difference in noisy conditions, and statistical linear approximation are along this direction. See S. Sagayama, Y. Yamaguchi, and S. Takahashi entitled “Jacobian Adaptation of noisy Speech Models," in Proceedings of IEEE Automatic Speech Recognition Workshop, pages 396-403, Santa Barbara, CA, USA, December 1997. IEEE Signal Processing Society. Also see “Statistical Linear Approximation for Environment Compensation” of N.S. Kim, IEEE Signal Processing Letters, 5(1): 8-10, January 1998.
- MLLR Maximum Likelihood Linear Regression
- a speech recognizer according to the present invention.
- the speech is applied to recognizer 11.
- the speech is compared to Hidden Markov Models (HMM) 13 to recognize the text.
- HMM Hidden Markov Models
- the models initially provided on those with speech recorded in a quiet environment and the microphone of good quality.
- a speech model set is provided using statistics about the noise and speech.
- a low computation cost method integrates both PMC and CMN.
- the first Step 1 is to start with HMM models trained on clean speech, with cepstral mean normalization. We modify these models to get models to compensate for channel/microphone distortion (convolutive distortion) and simultaneous background noise (additive distortion).
- convolutive distortion convolutive distortion
- additive distortion simultaneous background noise
- For an HMM model we have a lot of parameter but only change one subset of the parameters and that is mean vectors m p,j,k .
- the mean vectors m p,j,k of the original model space is modified where p is the index of the Probability Density Function (PDF), j is the state and k is the mixing component.
- PDF Probability Density Function
- the second Step 2 is to calculate which is the mean mel-scaled cesptrum coefficients (MFCC) vector over the trained database. Scan all data and calculate the mean to get b.
- MFCC mean mel-scaled cesptrum coefficients
- the fourth Step 4 is for a given input test utterance, an estimate of the background noise vector X ⁇ is calculated.
- Step 5 we calculate the mean vectors adapted to the noise X ⁇ using equation 4.
- m ⁇ p , j , k I D F T ( D F T ( m ⁇ p , j , k ) ⁇ D F T ( X ⁇ ) ) .
- DFT and IDFT are, respectively, the DFT and inverse DFT operation
- m ⁇ p,j,k is the noise compensated mean vector.
- Equation 4 involves several operators.
- DFT is Discrete Fourier Transform and IDFT is the Inverse Discrete Fourier Transform.
- the ⁇ is an operation with two vectors.
- a ⁇ B C. How ⁇ is defined, we look at equations 2 and 3.
- Equation 2 says the operation ⁇ operates on two vectors u and v and the result is a vector of D dimension or [ w 1 l , w 2 l , ... w D l ] T where T is the transposition. We take the two vectors and produce another vector. We need to specify each element in the resultant vector. Equation 3 says that the jth element in that vector ( w j l ) is defined by the exponential of the element of u added to the exponential of the jth element of v and take the log of the combination of the exponential of u added to the exponential of the jth element of v. This completes the definition of Equation 4.
- Equation 7 shows that b ⁇ can be worked out analytically, and it is not necessary to do the physically generation and integration.
- the final result is Equation 7 which is the integration into several sums. Sums over probability density functions and the sum over states and sum over mixing components. Then you have several quantities.
- the P H is the probability of having the PDF index.
- the P J given H is the probability of being in the state if given the PDFp.
- the next is the probability the mixing component p,j given we have the PDF index.
- This operation to calculate the b ⁇ can be simplified with three approximations.
- the first one uses equal probabilities for P H ( p ) or constraint C.
- CMN model m p , j , k , t T is then used with CMN on noisy speech.
- b ⁇ is a function of both channel and background noise in all above cases. In other words, in presence of noise, there is no guarantee that the channel will be removed by such a vector, as is for CMN.
- WAVES database containing hands-free recordings was used, which consists of three recording sessions: parked-trn (car parked, engine off), parked (car parked, engine off), and city-driving (car driven on a stop and go basis).
- each session 20 speakers (10 male) read 40 sentences each, giving 800 utterances.
- Each sentence is either 10, 7 or 4 digit sequence, with equal probabilities.
- the database is sampled at 8kHz, with MFCC analysis frame rate of 20ms.
- Feature vector consists of 10 static and 10 dynamic coefficients.
- HMMs used in all experiments are trained in TIDIGITS clean speech data. Utterance-based cepstral mean normalization is used. The HMMs contain 1957 mean vectors, and 270 diagonal variances. Evaluated on TIGIDIT test set, the recognizer gives 0.36% word error rate.
- the variances of the Gaussian PDFs can be MAP adapted to some slightly noisy data, e.g. WAVES parked_eval data. Such adaptation will not affect recognition of clean speech, but will reduce variance mismatch between HMMs the noisy speech.
- the new algorithm is referred to as JAC (joint compensation of additive noise and convolutive distortion).
- Table 1 word error rate (WER) (%) as function of driving conditions and compensations methods
- WER word error rate
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Circuit For Audible Band Transducer (AREA)
- Time-Division Multiplex Systems (AREA)
- Complex Calculations (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Claims (12)
- Procédé de modification de modèles cachés de Markov (HMM) originaux entraînés sur une parole pure avec une normalisation moyenne cepstrale pour obtenir des modèles adaptés qui compensent la distorsion convolutive et le bruit additif simultanés comprenant les étapes de :pour chaque énoncé vocal :calcul (2) d'un vecteur de coefficients cepstraux en l'échelle mel moyen (MFCC) b sur la base de données pure ;addition (3) du vecteur MFCC moyen b aux vecteurs moyens m p,j,k des modèles HMM originaux, où p est l'indice de la fonction de densité de probabilité (PDF), j est l'état et k la composante de mélange, pour obtenir m̅ p,j,k ;pour l'énoncé vocal donné, calcul (4) d'une estimation du vecteur de bruit de fond X̃ ;calcul (5) de vecteurs moyens de modèle adaptés au bruit m̂ p,J,k = IDFT(DFT(m̅ p,j,k ) ⊕ DFT(X̃)), où la transformée de Fourier discrète inverse est constituée de la combinaison ⊕ de la transformée de Fourier discrète des vecteurs moyens m̅ p,j,k et de la transformée de Fourier discrète du bruit estimé X̃ l'opérateur de combinaison ⊕ étant défini par wu l u l ⊕ v l = [w 1 l ,w 2 l ,...,w D l ] T avec u l [u 1 l ,u 2 l ,...,u D l ] T , v l [v 1 l ,v 2 l ,...,v D l ] T et W l j = log(exp(u j l ) + exp(v j̅ l )) ; etcalcul (6) d'un vecteur moyen b̂ des données bruitées sur l'espace vocal bruité, et élimination (7) du vecteur moyen b̂ des données bruitées des vecteurs moyens de modèle adaptés au bruit m̂ p,j,k pour obtenir les modèles adaptés.
- Procédé selon la revendication 1 dans lequel l'étape de calcul du vecteur moyen b des données bruitées sur l'espace vocal bruité calcule le vecteur en utilisant des statistiques de modèle bruité en utilisant :b̂ = ∑ p ∑ j ∑ k P H , (p)P J|H , (j|p)P K|H,J (k|p,j)m̂ p,j,k où H est la variable désignant l'indice de PDF, J est la variable pour l'indice d'état, et K est la variable pour l'indice de composante de mélange.
- Procédé de reconnaissance vocale avec compensation de la distorsion convolutive et du bruit additif simultanés, comprenant les étapes de :fourniture de modèles cachés de Markov (HMM) entraînés sur une parole pure avec normalisation moyenne cepstrale ; et modification desdits modèles par le procédé de l'une quelconque des revendications précédentes.
- Dispositif de reconnaissance vocale avec compensation de la distorsion convolutive et du bruit additif simultanés comprenant en combinaison :des moyens pour générer des modèles cachés de Markov (HMM) adaptés en modifiant des modèles HMM originaux entraînés sur une parole pure avec normalisation moyenne cepstrale, lesdits moyens de génération comprenant :des moyens pour, pour chaque énoncé vocal,calculer un vecteur de coefficients cepstraux à échelle de mel moyen (MFCC) b sur la base de données pure ;des moyens pour additionner le vecteur MFCC moyen b aux vecteurs moyens m p,j,k des modèles HMM originaux, où p est l'indice de la fonction de densité de probabilité (PDF), j est l'état et k est la composante de mélange, pour obtenir m̅ p,j,k ;des moyens pour, pour l'énoncé vocal donné, calculer une estimation du vecteur de bruit de fond X̃ ;des moyens pour calculer des vecteurs moyens de modèle adaptés au bruit m̂ p,j,k = IDFT(DFT(m̅ p,j,k ) ⊕ DFT(X̃)), où la transformée de Fourier discrète inverse est constituée de la combinaison ⊕ de la transformée de Fourier discrète des vecteurs moyens m̅ p,j,k et de la transformée de Fourier discrète du bruit estimé X̃, l'opérateur de combinaison ⊕ étant défini par w l u l ⊕ v l = [w1 l , ,w2 l ,...,wD l ] T avec u l [u 1 l , u 2 l , ,...,u D l ] T , v l [V l 1 , V l 2 ,..., V l D] T et w l j = log(exp(u l j ) + exp(v l j )) ; etdes moyens pour calculer le vecteur moyen b̂ des données bruitées sur l'espace vocal bruité, et des moyens pour éliminer le vecteur moyen b̂ des données bruitées des vecteurs moyens de modèle adaptés au bruit m̂ p,j,k pour obtenir les modèles adaptés ; etdes moyens pour comparer les modèles adaptés à l'énoncé vocal pour reconnaître la parole en entrée.
- Dispositif de reconnaissance selon la revendication 7 dans lequel les moyens pour calculer le vecteur moyen b̂ des données bruitées sur l'espace vocal bruité calculent le vecteur en utilisant des statistiques de modèle bruité en utilisant : b̂ = ∑ p ∑ j ∑ k P H (P)P J|H (j|p)P K|H,J (k|p,j) p,j,k
où H est la variable désignant l'indice de PDF, J est la variable pour l'indice d'état, et K est la variable pour l'indice de composante de mélange. - Procédé de reconnaissance vocale avec compensation simultanée de la distorsion convolutive et du bruit additif comprenant les étapes de :modification des modèles HMM par un procédé selon l'une quelconque des revendications 1 à 5.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US27548701P | 2001-03-14 | 2001-03-14 | |
US275487P | 2001-03-14 |
Publications (3)
Publication Number | Publication Date |
---|---|
EP1241662A2 EP1241662A2 (fr) | 2002-09-18 |
EP1241662A3 EP1241662A3 (fr) | 2004-02-18 |
EP1241662B1 true EP1241662B1 (fr) | 2006-06-21 |
Family
ID=23052506
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP02100251A Expired - Fee Related EP1241662B1 (fr) | 2001-03-14 | 2002-03-14 | Reconnaissance de la parole avec compensation de distorsion convolutif et de bruit additif |
Country Status (4)
Country | Link |
---|---|
US (1) | US7062433B2 (fr) |
EP (1) | EP1241662B1 (fr) |
JP (1) | JP2002311989A (fr) |
DE (1) | DE60212477T2 (fr) |
Families Citing this family (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6985858B2 (en) * | 2001-03-20 | 2006-01-10 | Microsoft Corporation | Method and apparatus for removing noise from feature vectors |
US20040148160A1 (en) * | 2003-01-23 | 2004-07-29 | Tenkasi Ramabadran | Method and apparatus for noise suppression within a distributed speech recognition system |
JP4357867B2 (ja) * | 2003-04-25 | 2009-11-04 | パイオニア株式会社 | 音声認識装置、音声認識方法、並びに、音声認識プログラムおよびそれを記録した記録媒体 |
JP4682154B2 (ja) * | 2004-01-12 | 2011-05-11 | ヴォイス シグナル テクノロジーズ インコーポレーティッド | 自動音声認識チャンネルの正規化 |
US7236930B2 (en) * | 2004-04-12 | 2007-06-26 | Texas Instruments Incorporated | Method to extend operating range of joint additive and convolutive compensating algorithms |
JP2006084732A (ja) * | 2004-09-15 | 2006-03-30 | Univ Of Tokyo | 多項式近似に基づく雑音下音声認識のためのモデル適応法 |
US20070033027A1 (en) * | 2005-08-03 | 2007-02-08 | Texas Instruments, Incorporated | Systems and methods employing stochastic bias compensation and bayesian joint additive/convolutive compensation in automatic speech recognition |
US7877255B2 (en) * | 2006-03-31 | 2011-01-25 | Voice Signal Technologies, Inc. | Speech recognition using channel verification |
CN1897109B (zh) * | 2006-06-01 | 2010-05-12 | 电子科技大学 | 一种基于mfcc的单一音频信号识别方法 |
US7664643B2 (en) * | 2006-08-25 | 2010-02-16 | International Business Machines Corporation | System and method for speech separation and multi-talker speech recognition |
CN101030369B (zh) * | 2007-03-30 | 2011-06-29 | 清华大学 | 基于子词隐含马尔可夫模型的嵌入式语音识别方法 |
US8180637B2 (en) * | 2007-12-03 | 2012-05-15 | Microsoft Corporation | High performance HMM adaptation with joint compensation of additive and convolutive distortions |
WO2009078093A1 (fr) | 2007-12-18 | 2009-06-25 | Fujitsu Limited | Procédé de détection de section non-parole et dispositif de détection de section non-parole |
US8306817B2 (en) * | 2008-01-08 | 2012-11-06 | Microsoft Corporation | Speech recognition with non-linear noise reduction on Mel-frequency cepstra |
US8145488B2 (en) * | 2008-09-16 | 2012-03-27 | Microsoft Corporation | Parameter clustering and sharing for variable-parameter hidden markov models |
US8214215B2 (en) * | 2008-09-24 | 2012-07-03 | Microsoft Corporation | Phase sensitive model adaptation for noisy speech recognition |
EP2182512A1 (fr) * | 2008-10-29 | 2010-05-05 | BRITISH TELECOMMUNICATIONS public limited company | Vérification du locuteur |
US8639502B1 (en) | 2009-02-16 | 2014-01-28 | Arrowhead Center, Inc. | Speaker model-based speech enhancement system |
CN103811008A (zh) * | 2012-11-08 | 2014-05-21 | 中国移动通信集团上海有限公司 | 一种音频内容识别方法和装置 |
US9489965B2 (en) * | 2013-03-15 | 2016-11-08 | Sri International | Method and apparatus for acoustic signal characterization |
CN106057195A (zh) * | 2016-05-25 | 2016-10-26 | 东华大学 | 一种基于嵌入式音频识别的无人机探测系统 |
US10720165B2 (en) * | 2017-01-23 | 2020-07-21 | Qualcomm Incorporated | Keyword voice authentication |
US20210201928A1 (en) * | 2019-12-31 | 2021-07-01 | Knowles Electronics, Llc | Integrated speech enhancement for voice trigger application |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5450522A (en) * | 1991-08-19 | 1995-09-12 | U S West Advanced Technologies, Inc. | Auditory model for parametrization of speech |
US5924065A (en) * | 1997-06-16 | 1999-07-13 | Digital Equipment Corporation | Environmently compensated speech processing |
US6529872B1 (en) * | 2000-04-18 | 2003-03-04 | Matsushita Electric Industrial Co., Ltd. | Method for noise adaptation in automatic speech recognition using transformed matrices |
US6912497B2 (en) * | 2001-03-28 | 2005-06-28 | Texas Instruments Incorporated | Calibration of speech data acquisition path |
-
2002
- 2002-01-18 US US10/051,640 patent/US7062433B2/en not_active Expired - Lifetime
- 2002-03-13 JP JP2002067939A patent/JP2002311989A/ja active Pending
- 2002-03-14 DE DE60212477T patent/DE60212477T2/de not_active Expired - Lifetime
- 2002-03-14 EP EP02100251A patent/EP1241662B1/fr not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
US7062433B2 (en) | 2006-06-13 |
US20020173959A1 (en) | 2002-11-21 |
DE60212477D1 (de) | 2006-08-03 |
JP2002311989A (ja) | 2002-10-25 |
DE60212477T2 (de) | 2007-07-05 |
EP1241662A3 (fr) | 2004-02-18 |
EP1241662A2 (fr) | 2002-09-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP1241662B1 (fr) | Reconnaissance de la parole avec compensation de distorsion convolutif et de bruit additif | |
US7165028B2 (en) | Method of speech recognition resistant to convolutive distortion and additive distortion | |
Segura et al. | Model-based compensation of the additive noise for continuous speech recognition. experiments using the Aurora II database and tasks. | |
US7269555B2 (en) | Unsupervised incremental adaptation using maximum likelihood spectral transformation | |
US6980952B1 (en) | Source normalization training for HMM modeling of speech | |
Ephraim et al. | On second-order statistics and linear estimation of cepstral coefficients | |
US7571095B2 (en) | Method and apparatus for recognizing speech in a noisy environment | |
US20020165712A1 (en) | Method and apparatus for feature domain joint channel and additive noise compensation | |
Huang et al. | An energy-constrained signal subspace method for speech enhancement and recognition in white and colored noises | |
Gong | A method of joint compensation of additive and convolutive distortions for speaker-independent speech recognition | |
Buera et al. | Cepstral vector normalization based on stereo data for robust speech recognition | |
US6633843B2 (en) | Log-spectral compensation of PMC Gaussian mean vectors for noisy speech recognition using log-max assumption | |
US6199041B1 (en) | System and method for sampling rate transformation in speech recognition | |
US7236930B2 (en) | Method to extend operating range of joint additive and convolutive compensating algorithms | |
Hansen et al. | Robust speech recognition in noise: an evaluation using the spine corpus | |
Nisa et al. | The speech signal enhancement approach with multiple sub-frames analysis for complex magnitude and phase spectrum recompense | |
US7225124B2 (en) | Methods and apparatus for multiple source signal separation | |
Torre et al. | On the comparison of front-ends for robust speech recognition in car environments | |
Kim et al. | Feature compensation scheme based on parallel combined mixture model. | |
Menéndez-Pidal et al. | Compensation of channel and noise distortions combining normalization and speech enhancement techniques | |
Yamamoto et al. | Speech recognition under noisy environments using segmental unit input HMM | |
Bernard et al. | Can back-ends be more robust than front-ends? Investigation over the Aurora-2 database | |
Torre et al. | Compensation of noise effects for robust speech recognition in car environments | |
Chien et al. | Bayesian affine transformation of HMM parameters for instantaneous and supervised adaptation in telephone speech recognition. | |
Delcroix et al. | Combined static and dynamic variance adaptation for efficient interconnection of speech enhancement pre-processor with speech recognizer |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE TR |
|
AX | Request for extension of the european patent |
Free format text: AL;LT;LV;MK;RO;SI |
|
PUAL | Search report despatched |
Free format text: ORIGINAL CODE: 0009013 |
|
AK | Designated contracting states |
Kind code of ref document: A3 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE TR |
|
AX | Request for extension of the european patent |
Extension state: AL LT LV MK RO SI |
|
RIC1 | Information provided on ipc code assigned before grant |
Ipc: 7G 10L 15/20 A |
|
17P | Request for examination filed |
Effective date: 20040818 |
|
AKX | Designation fees paid |
Designated state(s): DE FR GB |
|
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
RTI1 | Title (correction) |
Free format text: SPEECH RECOGNITION WITH COMPENSATION FOR BOTH CONVOLUTIVE DISTORTION AND ADDITIVE NOISE |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): DE FR GB |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: FG4D |
|
REF | Corresponds to: |
Ref document number: 60212477 Country of ref document: DE Date of ref document: 20060803 Kind code of ref document: P |
|
ET | Fr: translation filed | ||
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
26N | No opposition filed |
Effective date: 20070322 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 15 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 16 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20170222 Year of fee payment: 16 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20170223 Year of fee payment: 16 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: DE Payment date: 20170331 Year of fee payment: 16 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R119 Ref document number: 60212477 Country of ref document: DE |
|
GBPC | Gb: european patent ceased through non-payment of renewal fee |
Effective date: 20180314 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20181002 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: GB Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180314 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: FR Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180331 |